Implicit Memory

Theoretical Issues

Stephan Lewandowsky, John C. Dunn, Kim Kirsner

• Publisher: Psychology Press
• Year: 2014
• ISBN:
  9781317784937
• Edition: 1
• Topic: Psychologie
• Subtopic: Geschichte & Theorie in der Psychologie

The first to focus exclusively on implicit memory research, this book documents the proceedings of a meeting held in Perth, Australia where leading researchers in the field exchanged ideas, data, and predictions about theoretical issues. In addition to reporting new information on a variety of topics, integrating previous findings, and proposing new theoretical approaches to implicit memory, the book also contains critical commentaries by highly regarded area specialists.

I Introduction

1 The Role of Theory in Understanding Implicit Memory

Robert S. Lockhart
University of Toronto

Introduction

Growth in our understanding of human memory is much like William James's description of consciousness: a series of flights and perchings. The perchings are those settled periods, dull but undoubtedly necessary, during which there is general agreement over terminology and the major effort is spent in gathering data Flights on the other hand are periods of excitement and conceptual uncertainty, of new theories and novel experiments. The present volume makes it evident that memory theory is currently in full flight, prodded from its perch by the growing dissatisfaction with its own narrowness of focus and by the stimulation of its revived links with neuropsychology. One of the most intriguing problems to emerge from this period of flight centers on the phenomenon of implicit memory.

One index of the excitement and significance of any period of flight is the degree of conceptual turmoil. If the present analysis of implicit memory is any indication, present developments are wildly exciting and highly significant: current memory theory would seem to be suffering from a taxonomic crisis, if not a full-blown case of terminological chaos. The ubiquitous symptom is the dichotomy; the terms proceed two-by-two across the theoretical landscape like animals in search of an ark: implicit versus explicit, procedural versus declarative memory, abstractionists versus proceduralists, intentional versus unintentional, episodic versus semantic, conceptually-driven versus data-driven, and so on. It is not surprising, therefore, that most of the contributors spend considerable effort in staking out their cellular niche within this dichotomous environment. In this regard the contribution of Schacter, Bowers, and Booker is especially valuable in bringing a measure of order to the terminology, and their retreat into operationalism is probably a timely move. Also timely is the warning by Roediger, Srinivas, and Weldon that questions about the causes of dissociations posed in terms of these dichotomies may be too broad to yield clear answers. But in the midst of this profusion of theory there has emerged a body of data, reported and reviewed in this volume, that marks one of the most significant advances in our understanding of human memory that we have seen for some time. A question we might ask is whether the concomitant theory is a help, a hindrance, or merely an engaging distraction. It is certainly reasonable to ask whether devoting a volume to theoretical issues surrounding implicit memory might not be premature.

Do We Need so Much Theoretical Speculation?

It may seem to many observers of the field, and to some participants, that there is already too much speculative theory, and that the data are simply not rich enough to carry the theoretical weight demanded of them, or sufficiently decisive to settle the various theoretical debates to be found in the literature generally, and in this volume in particular. Or it might be argued that most of the theoretical formulations are so conceptually vague or obscure as to be impervious to data. Snodgrass for example, suggests that some aspects of the theoretical debate will simply grind to a halt when researchers realize that their resolution is impossible in principle, and in this I think she may be correct But does this prognosis mean that theoretical speculation is a waste of time, an idle distraction needed only to write acceptable introductions and discussion sections to journal articles?

To answer this question it is necessary to examine the function that theory construction might serve in increasing our understanding of a phenomenon such as implicit memory. In particular, we need to understand the interactive role of data and theory. It is usually thought that this role consists of the data deciding the fate of the theory by a process of conjecture and attempted refutation. But at this early stage in the history of cognitive psychology, when we can safely assume that all theories are far from the ultimate truth, the interactive relation of data and theory has less to do with the survival of the theory than it has with the survival of the data. The major function of theory should be to guide data gathering; it should be to ensure that the data are archival in the sense that their importance outlasts the life of the theory, and that the data are cumulative in the sense that experiments build a data base that any theory, present or future, must take into account.

This point of view is captured by Wittgenstein in the final proposition (6.54) of the Tractatus. Wittgenstein of course was referring to his preceding propositions, not to the relationship between data and theory, but his metaphor captures the point exactly:

My propositions serve as elucidations in the following way: anyone who understands me eventually recognizes them as nonsensical, when he has used them—as steps—to climb up beyond them. [He must, so to speak, throw away the ladder after he has climbed up it.] He must transcend these propositions, and then he will see the world aright

(Wittgenstein, 1921/1961, p. 74).

The question we might therefore ask of the theoretical speculations in this volume is whether (1) they have led to new experiments yielding archival data or (2) they are merely a way of gaining a post hoc structuring of the data. With a few possible exceptions I think (1) is the correct alternative. The data reported in the chapter by Lewandowsky, Kirsner, and Bainbridge, for example, provide a good illustration supporting this point, as do the data from many of the other chapters. Lewandowsky et al.'s experiments are motivated by the view that context effects would pose difficulties for all major current theoretical accounts of implicit memory, and they use their results to argue for a theory that posits multiple memory representations. I doubt whether this initial assumption is justified (they underestimate the flexibility of the theories as well as the ingenuity of their creators) and I will argue later in the chapter that their conclusion is also questionable. But I have no doubt that their theory of sense-specific activation has served a useful purpose—a Wittgenstein ladder—even if the theory should now be discarded; we have climbed a little higher and can see a little more clearly. Their data are archival in that it is difficult to believe that any complete account of implicit memory will be able to ignore them and their value is largely independent of the theory that prompted their collection. Only the judgment of history will tell us whether I am justified in holding such an optimistic view of the life-expectancy of these and other data reported in this volume. The optimism is not offered lightly; it is a sobering exercise to skim the journals devoted to human learning and memory, published between say 1950 and 1970, and to estimate the percentage of experimental results that hold any interest for, or exert any sustained influence on, contemporary memory theory.

My suggestion then is that although theoretical speculation such as the contrast between activation, procedural, and system theories may ultimately prove to be a spurious debate as Snodgrass suggests, the speculation cannot be accused of baking no bread, to use William James's expression, On the other hand, from the viewpoint of generating new data, I am less optimistic about the mathematical theories described in the chapters by Weber and Murdock and by Humphreys, Bain, and Burt, their precision and elegance notwithstanding. The function of their mathematical formalism would seem to be to account for existing data rather than to suggest new experiments and their success in doing so is their major achievement. Such post-hoc structuring of the data is an entirely justified goal, provided of course there exists enough data and it is of the right kind; every Kepler presupposes a Tycho Brahe, but the question is whether the time for a Kepler might not be a bit premature.

Implicit Memory, Imageless Thought

The term implicit memory reminds me of two much older expressions: unconscious inference and imageless thought. All three terms have a slightly odd ring about them in that each takes a cognitive process normally associated with conscious mental activity—remembering, reasoning, thinking—and forms a qualified concept by deleting the conscious component. This subtractive strategy, in which a more elementary phenomenon is expressed as a simplified version of a more complex one, might seem a strange style of characterization—rather like defining water as distilled orange juice, or a bicycle as a motorcycle without an engine. In feet it reveals a good deal about theoretical perspectives and biases. The pattern of qualification suggests that it is conscious mental activity that is the theoretical reference point. Why else should we be impressed by the fact that a response to a current stimulus can be influenced by a past event of which the subject is unaware? Isn't that how most organisms behave most of the time, indeed bow most organisms, including infants, behave all the time? It might be helpful to consider taking a different reference point, one that could be argued to be phylogenetically and ontogenetically more basic (see Sherry & Schacter, 1987). That is, we might regard implicit memory as the norm (and call it something like learning, perceptual fluency, or cognitive—Gibson-style—attunement, as has occasionally been suggested) and regard explicit memory to be the specialized capacity of a highly evolved species. Unfortunately this line of argument is not strongly represented in the present volume, although in his chapter, Parkin gives us some interesting evidence further supporting the argument for the ontogenetic priority of implicit memory (for example, see Lockhart, 1984; Nelson, 1988; Schacter & Moscovitch, 1984; Tulving, 1983).

As Jacoby (1988) has pointed out there are some interesting parallels between the current discussion of implicit memory and the debate that surrounded the alleged phenomenon of imageless thought studied so intensively during the first decade of this century at Külpe's laboratory in Würzburg. The basic demonstration of imageless thought was that performance on a task could be influenced by a prior stimulus (usually an instructional set) without the subject being aware of that prior stimulus while executing the task. The prior stimulus was said to establish a determining tendency, that set (primed?) the subject to respond in one way rather than another, but whose presence in consciousness was not a necessary (or even actual) part of the response process. H. J. Watt carried out the early experiments on imageless thought and in form they were very similar to contemporary studies of lexical access. Subjects (trained introspectionists of course) were presented with an instruction (e.g., superordinate), the stimulus word was presented (dog), and the time to respond (animal) was measured using the Hipp chronoscope along with the subjects' introspective report of the contents of their consciousness while making the response.

It is tempting to dismiss these experiments as the largely worthless efforts of experimentalists hopelessly committed to the introspective method. In fact, Watt's experiments, and the work that followed, made an amazing discovery, one so counter-intuitive that, judging from reactions to contemporary work in implicit memory, it is as counter-intuitive today as it was in 1904. Watt divided consciousness into four sequential periods: the preparatory period, the appearance of the stimulus word, the search for the response word, and the occurrence of the response. Boring (1950) points out that "every one had been expecting to find the key to thought in Watt's third period, the period of search for the word that would satisfy the conditions" (p. 404). Watt's discovery was that as far as conscious content is concerned, it was the first period, not the third, that mattered. Watt had demonstrated that even for the 'higher mental processes' a response to a stimulus could be determined by events that were not part of consciousness.

There are a number of important differences of course between these old experiments and contemporary studies of implicit memory. In the typical implicit memory experiment, subjects are unaware of the purpose of the prime; in the Würzburg experiments the instructional set was explicit so that the determining tendency is a type of intentional or conscious priming discussed by Humphreys, Bain, & Burt in their chapter, and of the five examples described by Schacter, Bowers, and Booker, the Würzburg task probably best matches the fourth. However the important point is that in influencing the subject's response to the presented test stimulus, the causal role of the priming stimulus was implicit in the sense that its effect was not mediated by its being consciously recollected.

Why was this result surprising? The counter-intuitive reaction seems to stem from an assumption, shared by the introspectionists, eschewed by the behaviorists, but revived by contemporary cognitivists, that the major determinants of 'higher-level' cognitive behavior are available to consciousness. This despite such warnings as Nisbett and Wilson's (1977) that such determinants might be unavailable to consciousness, or by Pylyshyn (1973) that what is available may not be what is important. Tulving (1989) has used the term concordance to describe this supposition of a simple correspondence between underlying process and the contents of consciousness and has also called it into question. However, the conclusion that Tulving draws from this observation is that, far from being irrelevant, consciousness is an important object of study, precisely because it does not stand in any simple one-to-one (concordant) relation with psychological process. I think Tul ving's argument is a sound one and that consciousness will continue to haunt discussions of memory even if its ghostly presence remains unnoticed and unmentioned. The exorcism attempted by Schacter, Bowers, and Bookers is a reasonable application of operationalism provided it is understood, as I think they intended it to be, as a temporary measure (Wittgenstein's ladder) needed to get on with the job of building a conceptually unambiguous data base. But as their own examples demonstrate, it is impossible to describe the range of memory performance without an appeal to consciousness. I think this conclusion holds even if one accepts Hirst's coherence model which relegates consciousness to a peripheral role.

Is Implicit Memory Merely an Example of Automaticity?

Perhaps our willingness to be impressed or even surprised by the influence of implicit memory is quite misplaced, and the entire phenomenon is a simple matter of automatized responding. After all, in performing a well-practiced motor skill an absence of concordance would not be considered unusual. It might be surprising if a person could not recollect yesterday's hour-long practice session on the pursuit-rotor task, but no one believes that the increased level of performance gained from that session depends in any direct way on the subject's ability to recollect it while consciously performing the task.

But responses id the studies of imageless thought or implicit memory are not merely the running off of an automatized or overlearned skill. Such overlearning is undoubtedly an important, perhaps essential, ingredient as suggested by results such as those of Graf and Schacter (1987) with 'new associations'. The important point is that the 'triggered' response can be altered and controlled by a single preceding episode. It is not just that animal is a well-practiced response to the stimulus word dog but that its emission as a response can be controlled—turned on or off—by a prior instructional stimulus. Thus which of the two responses, animal or beagle, the subject gives to the stimulus word dog, is determined by the prior instruction (superordinate or subordinate), not exclusively by the relative degree of practice or associative strength between the stimulus and the response. The claim of the Würzburgers was that whatever caused this differential responding had no conscious component. They described the process as being 'imageless'; the current term is 'implicit'.

This capacity of prior instructions to 'set' a mode of automatized responding is a phenomenon that has been given insufficient attention in contemporary work. Bransford and Franks (1976) used a 'stage-setting' metaphor to describe the general way in which the cognitive system can be pre-tuned; Humphreys, Bain, and Burt in their chapter in this volume use the term 'potentiate' to refer to the fact that instructions can determine the form of memory retrieval in which the subject engages. One of the very impressive aspects of their model is its success in tackling the apparently easy (but in fact extremely difficult) matter, raised in the paragraph above, of how the system can preset itself to override, rather than be overwhelmed by, pre-experimental associations. This problem is one worth continued attention. The concept of potentiation raises the general question of the selectivity of priming, not just for the instructionally explicit 'stage-setting' priming described above, but also for the more usual instructionally covert or incidental variety. In our normal waking state we are subject to a constant stream of incoming stimuli. Given that priming can be quite incidental, it would seem to follow that we are in a stale of being continuously primed by our everyday environment. This constant modulation of our cognitive sensitivities raises some additional questions. We might ask about the functional significance, if any, of all this priming. Does priming have direct functional significance—preparing us to respond more rapidly or with greater sensitivity to the second occurrence of a stimulus? Or is priming a gratuitous by-product of other, more fundamental, properties of the cognitive system? We might also ask if there are complementary inhibitory processes that, analogously to vision, sharpen cognitive contours and prevent the general overloading of the system. What dimensions of context are relevant to the capacity of the response to one stimulus to prime the response to another? That we already have some answers to these questions (especially to the last-mentioned) will be apparent to any reader of this volume, but it must be equally apparent that complete answers require more data.

Task and Process

A valuable contribution that is helpful in clarifying the profusion of theoretical dichotomies is the distinction between task and process made by Dunn and Kirsner. They coin the term transparency assumption to refer to the supposition that tasks of a particular type give a direct reflection of a particular underlying process. The operationalism of Schacter, Bowers, and Booker can be thought of as an effort after such transparency; after all, tasks are informative only if they are at least translucent and allow some glimpse, however shadowy, of the underlying process. Thus, as Dunn and Kirsner point out, the relation between task and process must be specified.

Two possible applications of Dunn and Kirsner's ideas are worth emphasizing. One is the possible role of implicit processes in explicit remembering. The other is the role of non-memorial component processes in episodic amnesia. On the first point, there has been considerable previous discussion of implicit memory as a component of recognition memory and the question is addressed by several contributors to this volume, especially Dunn and Kirsner and Hirst. Nilsson and Bäckman also discuss the matter in relation to subject performed tasks (SPTs). But what of the processes underlying various forms of recall? Nilsson and Bäckman suggest that the difference between SPTs and verbal tasks (VTs) is that for the former implicit memory components might be involved as well. But it is difficult to believe that the difference is a simple matter of implicit processes being present in SPTs and absent in VTs. Their attempts to identify the implicit processes in SPTs might equally well be appl...

Table of contents

• Cover Page
• Half Title page
• Title Page
• Copyright Page
• Contents
• Preface
• Participants of Implicit Memory: Theoretical Issues
• Part I: Introduction
• Part II: Characterizing Implicit Memory
• Part III: Theories and Models
• Part IV: Processes and Representations
• Part V: Development and Learning
• Part VI: Comment
• References
• Author Index
• Subject Index